PII RedactorDOCX Engine
Engine Gateway
Deterministic Word Document Sanitization

Zero-Knowledge PII Redaction for Microsoft Word

Sanitize sensitive data across structured text, contextual NLP entities, and embedded image scans without losing Word formatting, tables, shading, or custom fonts.

10 PII Categories
Direct XML Byte Splicing
0-Byte Ephemeral RAM Stream
Luhn & SSA Validated

Document Redaction Workspace

OPENXML .DOCX
or drag and drop

Supports Microsoft Word (.docx) packages up to 50MB

XML Text RunsOCR Embedded ImagesFormatting Intact
Social Security NumbersStructured (Go)

Matches SSN syntax, validates SSA area codes (no 000, 666, 900-999), and swaps with deterministic pseudonyms.

RAW:123-45-6789
FAKE:987-65-4321
SSA Allocation Rules Check
Credit & Debit CardsStructured (Go)

Rejects arbitrary 16-digit order numbers that fail Luhn. Valid cards are deterministically replaced.

RAW:4532 0150 2831 9283
FAKE:4111 2948 1029 8472
Luhn Checksum Algorithm (13–19 digits)
Phone NumbersStructured (Go)

Validates national and international area codes. Non-conforming numbers are ignored.

RAW:+1 (415) 555-0192
FAKE:+1 (415) 555-0188
Google libphonenumber Parser
Email AddressesStructured (Go)

Full mailbox and domain validation. Substitutes realistic pseudonym email formatting.

RAW:j.vance@acme-health.org
FAKE:m.sterling@apex-clinical.org
RFC 5322 Syntax Rules
IP AddressesStructured (Go)

Detects and validates IPv4 and IPv6 addresses, preserving valid subnet notation.

RAW:192.168.1.104 / 2001:db8::1
FAKE:10.0.4.82 / 2001:db8:85a3::1
net.ParseIP Subnet Parser
Dates of BirthStructured (Go)

Distinguishes birthdates from invoice and contract dates using calendar layouts and age bounds.

RAW:November 23, 1984
FAKE:April 16, 1982
time.Parse + Lifespan Validity
Full Person NamesNLP (Python)

spaCy NER model filtered to multi-token names to eliminate false positives on common solitary nouns.

RAW:Jonathan Vance
FAKE:Marcus Sterling
spaCy PERSON NER (2+ Tokens Required)
Organizations & CompaniesNLP (Python)

Dual-signal matching: contextual NER combined with entity suffix patterns (Inc, LLC, Corp, GmbH).

RAW:Metro Health Partners LLC
FAKE:Beacon Health Network LLC
spaCy ORG NER + Legal Suffixes
Physical Street AddressesNLP (Python)

Matches street number, name, unit/suite suffixes, and postal boundary heuristics.

RAW:742 Evergreen Terrace, Suite 400
FAKE:104 Meadow Lane, Suite 210
Street-Type Pattern Rules
Embedded Image OCRVisual OCR

Extracts ZIP media, runs OCR character recognition, and splices masked PNG/JPEGs back into the DOCX archive.

RAW:Scanned State ID / Photo
FAKE:Blackout Bounding Box Mask
Tesseract OCR + Pillow Masking